Converts an NDF to a text file NDF2ASCII
"Data", "Quality" or "Variance". ["Data"] TRUE, any FITS extension is written to start of the output file, unless there is no
extension whereupon a minimal FITS header is written to the ASCII file. [FALSE] TRUE, the output file allocates a fixed number of
characters per data value. The number of characters chosen is the minimum that prevents any
loss of precision, and hence is dependent on the data type of the NDF array. These widths
in characters for each HDS data type are as follows: _UBYTE, 3; _BYTE, 4; _UWORD, 5;
_WORD, 6; _INTEGER, 11; _INT64, 20; _REAL, 16; and _DOUBLE, 24. The record length is the
product of the number of characters per value plus one (for a delimiting space), times
the number of values per record given by parameter NOPEREC, up to a maximum of
512.
When FIXED is FALSE, data values are packed as efficiently as possible within each record. The length
of each record is given by Parameter RECLEN. [FALSE]
TRUE. It
should be positive on UNIX platforms. The suggested default is the current value, or 8 when there is
not one. The upper limit is given by 512 divided by the number of characters per value plus 1 (see
Parameter FIXED). FALSE and will default to the current value, or 132 if there is no current value. [] cluster.dat. The maximum recordlength of cluster.dat is
132 bytes, and the data values are packed into these records as efficiently as possible. cluster.dat. The maximum recordlength of cluster.dat is 132 bytes, and the
variance values are packed into these records as efficiently as possible. cluster.dat. There are twelve data values per record in cluster.dat. ndf234.dat. The maximum recordlength of ndf234.dat is 80 bytes, and
the data values are packed into these records as efficiently as possible. If there is a FITS
extension, it is copied to ndf234.dat with substitution of certain keywords, otherwise a
minimal FITS header is produced. the NDF array as selected by COMP is written to the ASCII file in records following an optional
header. When FIXED is FALSE all records are padded out to the recordlength.
The NDF array elements are written in Fortran order, i.e. the first dimension varies fastest, followed by the second dimension and so on. For example, a 2x2x2-element cube’s indices will appear in the order (1,1,1), (2,1,1), (1,2,1), (2,2,1), (1,1,2), (2,1,2), (1,2,2), (2,2,2).
HISTORY is not propagated.
ORIGIN information is lost.
When a header is to be made, it is composed of FITS-like card images as follows:
The number of dimensions of the data array is written to the keyword NAXIS, and the actual dimensions to NAXIS1, NAXIS2 etc. as appropriate.
If the NDF contains any linear axis structures the information necessary to generate these structures is written to the FITS-like headers. For example, if a linear AXIS(1) structure exists in the input NDF the value of the first data point is stored with the keyword CRVAL1, and the incremental value between successive axis data is stored in keyword CDELT1. By definition the reference pixel is 1.0 and is stored in keyword CRPIX1. If there is an axis label it is written to keyword CTYPE1, and axis unit is written to CUNIT1. (Similarly for AXIS(2) structures etc.) FITS does not have a standard method of storing axis widths and variances, so these NDF components will not be propagated to the header. Non-linear axis data arrays cannot be represented by CRVALn and CDELTn, and must be ignored.
If the input NDF contains TITLE, LABEL, or UNITS components these are stored with the keywords TITLE, LABEL, or BUNIT respectively.
If the input NDF contains a FITS extension, the FITS items may be written to the FITS-like header, with the following exceptions:
An extra header record with keyword UNSIGNED and logical value T is added when the array data type is one of the HDS unsigned integer types. This is done because standard FITS does not support unsigned integers, and allows (in conjunction with BITPIX) applications reading the unformatted file to determine the data type of the array.
The last header record card will be the standard FITS END.
Other extensions are not propagated.
All non-complex numeric data types are supported.
The value of bad pixels is not written to a FITS-like header record with keyword BLANK.